Feature extraction

Open Theon767 opened this issue 9 months ago • 0 comments

Hello, thanks to your great work. Right now i'm trying to reproduce the training process of your work and i'm stuck in feature extractions.

When i try to run python tools/extract_features.py ${CFG_PATH} --save-dir ${OUT_DIR} --checkpoint ${CKPT_PATH} --projections-dir ${PROJ_DIR} --complete, i got size mismatch for decode_head.text_embeddings: copying a param with shape torch.Size([171, 512]) from checkpoint, the shape in current model is torch.Size([28, 512]).

What do you mean by "We do not use these weights during feature extraction.“， and how can i correctly extract features?

Best wishes

Apr 16 '25 02:04 Theon767